Algorithms for the People: Democracy in the Age of AI

Abstract

Our society is being transformed by prediction tools like artificial intelligence and machine learning. And yet, we find ourselves chasing tech companies whose AI systems we know nothing about, condemning algorithms that entrench racial inequality in the criminal justice system, and struggling to hold accountable those who build and use the predictive technologies reshaping our world, whether welfare agencies or police forces, Facebook or Google. Algorithms for the People deploys the tools of political theory to flip the narrative around technology governance. Instead of exploring the impact of technology on democracy, this dissertation explores what the pursuit of a resilient and healthy democracy should mean for how we govern technology, connecting debates about AI ethics to ancient questions of justice and democracy. The dissertation develops an accessible and systematic account of what technologies like AI and machine learning are, why they are political, and how the institutions that deploy them should be regulated – a political theory of machine learning. The dissertation brings together two debates that are too often disconnected: debates about algorithmic fairness and discrimination and debates about the regulation of Facebook and Google. By exploring the political problems posed by the design and use of machine learning systems in these two contexts, the dissertation shows how technology regulation and democratic reform are connected, setting out a vision for regulating machine learning that places the flourishing of democracy at its heart.